Earth-abundant water-splitting catalysts coupled to silicon solar
cells for solar-to-fuels conversion
(Article begins on next page)
The Harvard community has made this article openly available.
Please share how this access benefits you. Your story matters.
Citation
No citation.
Accessed
February 17, 2015 1:19:21 AM EST
Citable Link
http://nrs.harvard.edu/urn-3:HUL.InstRepos:13070034
Terms of Use
This article was downloaded from Harvard University's DASH
repository, and is made available under the terms and conditions
applicable to Other Posted Material, as set forth at
http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAA
Earth–abundant water–splitting catalysts coupled to
silicon solar cells for solar–to–fuels conversion.
A dissertation presented by Casandra R. Cox to
The Department of Chemistry and Chemical Biology
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in the subject of
Chemistry Harvard University Cambridge, Massachusetts September 2014
© 2014 by Casandra R. Cox
Casandra R. Cox Daniel G. Nocera
Earth–abundant water–splitting catalysts coupled to silicon
solar cells for solar–to–fuels conversion.
Abstract
Direct solar–to–fuels conversion can be achieved by coupling
semiconductors with water–splitting catalysts. A 10% or higher solar to fuels conversion is minimally necessary for the realization of a robust future technology. Many water–splitting devices have been proposed but due to expensive designs and/or materials, none have demonstrated the necessary efficiency at low–cost that is a requisite for large–scale implementation. In this thesis, a modular approach is used to couple water–splitting catalysts with crystalline silicon (c–Si) photovoltaics, with ultimate goal of demonstrating a stand–alone and direct solar-‐to-‐fuels water– splitting device comprising all non–precious, technology ready, materials.
Since the oxygen evolution reaction is the key efficiency–limiting step for water–splitting, we first focus on directly interfacing oxygen evolution catalysts with c–Si photovoltaics. Due to the instability of silicon under oxidizing conditions, a protective interface between the PV and OER catalyst is required. This coupling of catalyst to Si semiconductor thus requires optimization of two interfaces: the silicon|protective layer interface; and, the protective layer|catalyst interface. A modular approach allows for the independent optimization and analysis of these two interfaces.
A stand–alone water–splitting device based on c–Si is created by connecting multiple single junction c-‐Si solar cells in series. Steady–state equivalent circuit analysis allows for a targeted solar–to–fuels efficiency to be designed within a predictive framework for a series–connected c–Si solar cells and earth–abundant water–splitting catalysts operating at neutral pH. Guided by simulation and modeling, a completely modular, stand–alone water–splitting device possessing a 10% SFE is demonstrated. Importantly, the modular approach enables facile characterization and trouble–shooting for each component of the solar water– splitting device. Finally, as direct solar water–splitting is far from a mature
technology, alternative concepts are presented for the future design and integration of solar water–splitting devices based on all earth–abundant materials.
Table of Contents
Title page i Copyright page ii Abstract iii
Table of Contents v
List of Figures viii
List of Tables xiii
List of Abbreviations xiv
Acknowledgments xvi–xix 1. Chapter 1–Introduction 1
1.1.The need for clean energy 2
1.2.Renewable energy 3
1.3.Capture of solar power and conversion to electrical power 4 1.4.Conversion of electrical power into fuels 5
1.5.Photoelectrochemical water–splitting 8
1.5.1. Buried–junction PEC requirements 9
1.5.2. Buried–junction PEC devices 11
1.6.Crystalline Silicon 11
1.7.Earth–abundant water–splitting catalysts 13
1.8.Overview 14
2. Chapter 2–Interfaces between crystalline silicon
solar cells and water–oxidation catalysts 24
2.1.Introduction 25
2.2.Results 27
2.2.1. Optimization of OER–catalyst functionalized
silicon solar cells 32
2.3.Discussion 37
2.4.Conclusion 39
2.5.Experimental 40
2.6.References 46
3. Chapter 3–Modeling a coupled photovoltaic electrochemical
devices using steady–state equivalent circuit analysis 50
3.1.Introduction 51
3.2.Efficiency considerations 52
3.3.Steady–state equivalent circuit analysis 57
3.4.Results and Discussion 60
3.4.1. Impact of ηPV on SFE 61
3.4.2. Impact of ηEC efficiency on SFE 63
3.5.Model validation 65
3.6.Conclusion 67
3.7.Experimental 68
4. Chapter 4–10% solar–to–fuels efficiency with non–precious materials 73 4.1.Introduction 74 4.2.Results 75 4.2.1. Device integration 80 4.3.Discussion 84 4.4.Conclusion 90 4.5.Experimental 91 4.6.References 95
5. Chapter 5–Future directions 98
5.1.Introduction 99
5.2.Alternative PV materials 99
5.3.Alternative catalyst deposition methods 101
5.4.Cell design 106 5.5.Conclusion 108 5.6.Experimental 108 5.7.References 110
List of Figures
Figure 1.1 Schematic showing (1) solar capture of solar energy by a photovoltaic device, (2) conversion of solar photons into a wireless current, and (3) storage via breaking the bonds of H2O to make H2 which can be used as a fuel. Adapted from ref.
5. 3
Figure 1.2 Solar irradiance at the surface of the Earth. The band–gap of silicon is overlaid as an example showing that photons absorbed at the band–gap can be converted and those absorbed above the band–gap are wasted as heat. 4
Figure 1.3 Qualitative schematic of an n–type semiconductor/electrolyte junction
for photoelectrochemical water–splitting. 9
Figure 1.4 Qualitative schematic of a buried–junction photovoltaic interfaced with water–splitting catalysts via Ohmic contacts for solar–water–splitting. 10
Figure 1.5 Chinese c–Si PV module prices since 2006. The data was recreated from
ref. 43. 12
Figure 1.6 Depictions of the molecular structure of our Mn, Co, and Ni water– oxidation catalysts. Reprinted with permission from Mike Huynh. 14
Figure 2.1 Schematic of the OER–catalyst functionalized silicon solar cell used in these studies. For electrochemical measurements, an external voltage may be applied to the contacts at either side of the cell with or without illumination. A. The solar cell is operating under reverse bias conditions with voltage applied to the front metal contacts on the n–side in the dark. B. The voltage can be applied directly to the protective–layer, in which case the PV is bypassed and the current–voltage characteristics are those of the OER–catalyst on an electrode. C. The solar–cell is illuminated with AM 1.5 illumination and the current–voltage behavior reflects the activity of the OER–catalyst functionalized solar cell. 27
Figure 2.2 CV curves of (top) npSi|ITO|CoPi and (bottom) npp+Si|ITO|CoPi in 0.1 M KPi electrolyte at pH 7 in the dark with Vappl through the n–side of the cell (▬▬, black), with Vappl thought the n–side under 1 sun AM 1.5 illumination (▬▬, green) , and in the dark with Vappl through the ITO layer bypassing the PV (▬▬, blue). Taken
from ref. 20. 28
Figure 2.3 Schematic showing the band-‐diagrams at the p-‐Si|ITO interface. A. Before contact. B. After equilibration of Fermi levels after interfacing p–Si with ITO. C. After equilibration of the Fermi levels after interfacing p+–Si with ITO. 30
Figure 2.4 Tafel plots for npp+Si|ITO|CoPi with potential applied to the metal front contact for measurements in dark (black squares, n), at 100 mW cm–2 (green squares, n), and 1000 mW cm–2 (orange red squares, n) illumination. The blue triangles (▲) correspond to a measurement in dark where the potential was applied through the ITO film at the back of the sample. Figure taken from ref 20. 31
Figure 2.5 Representative J–V curve for generation 2 npp+–Si solar cells used in this study in the dark (▬▬, black) and under AM 1.5 illumination (▬▬, blue). 33
Figure 2.6 Plane view SEM images of OER–catalyts deposited on surface–protected npp+Si|electrodes. From left to right A npp+|FTO|CoBi and B npp+Si|Ni|CoBi. 34
Figure 2.7 O2 production measured by a fluorescent sensor (▬▬, red) and the amount produced based on current passed assuming 100% Faradaic efficiency (▬▬, green) for (left) npp+–Si|FTO|CoBi and (right) npp+–Si|Ni|CoBi. 35
Figure 2.8 Tafel plots of (a) npp+Si|ITO|CoBi (b) npp+Si|FTO|CoBi and (c) npp+Si|Ni|CoBi. With the potential applied to the metal front contact for measurements in the dark (●), under 1 sun AM 1.5 illumination (●), and in the dark
with the potential applied through the protective coating at the back of the sample
(●). 36
Figure 2.9 Graph showing the variability in Tafel slope for various combinations of OER–catalyst functionalized c–Si solar cells. The red lines indicate the value based
on previously reported Tafel analysis. 38
Figure 3.1 Schematic of a wired and wireless PV–EC based on silicon solar cells. Regardless of the mode of coupling between the two, the equivalent circuit is
identical. 55
Figure 3.2 Block diagram for a photovoltaic (PV) powered electrochemical cell (EC), where direct electrical connection constrains JPV = JEC and VPV = VEC. 56
Figure 3.3 The generalized current density–voltage (J–V) diagram of a coupled PV– EC system where the point of intersection of the PV–curve (▬▬, blue) and EC– curve (▬▬, red) represents the operational point and SFE of the coupled PV–EC device. The SFE is maximized when the operating point is equal to PMAX. 57
Figure 3.4 Impact on the J–V curve for a PV due to shunt (▬▬,dark red) or series (▬▬,dark green) resistance compared to an ideal J–V curve (▬▬,dark blue). 58
Figure 3.5 Steady–state equivalent circuit of a PV–EC system. An applied voltage is incorporated to illustrate analysis of an externally assisted system. 60
Figure 3.6 Impact on SFE via improvement in PV efficiency compared to the baseline ηPV = 20% (–––––, grey dash). Given optimal coupling between the PV and EC components (Top) a higher relative SFE can be obtained by improving the JSC (▬▬, green) as opposed to the VOC (–––––, dashed green). Given poor coupling between the baseline PV and EC (bottom), only minor improvements in the SFE can
be obtained. 62
Figure 3.7 J–V curves of multiple series connected solar cells with ηPV = 20% (▬▬, grey) and EC curves (▬▬, dark blue). The number of solar cells required changes based on choice of catalyst which causes the EC curve to shift left or right and resistive losses due to RSOL cause the EC curve to tilt down. 63
Figure 3.8 Impact of solution resistance and EC parameters on SFE given ηPV = 20%. Case I EC parameters (▬▬, green) are based on utilizing the Co–OEC and Case II EC
(▬▬, navy) are based on utilizing the Ni–OEC. 65
Figure 3.9 Graphical demonstration of how the predictive analysis works for PV– assisted reactions, where the PV–curve (▬▬, blue) is based on the J–V characteristics of an in–house built single junction c–Si PV and the EC–curve (▬▬, red) is based on the CoBi water–oxidation catalyst operating in pH 9.2 solution. 66
Figure 3.10 Predicted Tafel behavior of a PV–assisted water oxidation system similar to the experiments described in Chapter 2. The electrical properties of the PV (shown in Fig. 3.8) and EC systems were measured independently (●, black
dots) and used to predict the coupled behavior (▬▬, black). The Tafel analysis of the PV–assisted photoanode (●, red dots) and predicted behavior match to within
10 mV. 67
Figure 4.1 Schematic of a PV–EC device based on series–connected single–junction c–Si solar cells and water–splitting catalyst. In this configuration the OER–catalyst is directly deposited on the back of the last solar cell in the stack. 76
Figure 4.2 Schematic of a PV–EC device used in these studies. In this modular configuration each component can be easily evaluated and replaced independently.
77
Figure 4.3 J–V curves of the individually measure PV and EC components making up the PV–EC device. The grey curves represent the J–V curves for the PV modules composed of either three (
––––
, grey–dashed) or four (▬▬, grey–solid) single– junction c–Si solar cells measure under AM 1.5 illumination. The red curves represent electrochemical load J–V curves using NiBi and NiMoZn catalysts, where the ideal EC curve (––––
, red–dashed) is based on previously reported Tafel analysis and the actual EC curve (▬▬, red) measured in a 2–electrode experiment (0.5M KBi / 0.5M K2SO4, pH 9.2). The point of intersection represents the JOP (●, orange circles) and the SFE of the coupled system. 78 Figure 4.4. Steady–state current voltage behavior for the NiBi operating in 0.5M KBi / 0.5 M K2SO4 pH 9.2 in H2 saturated solution (●) and in Ar saturated solution (▲).Since the voltage required to achieve a given current density under both conditions is almost identical indicates that the contribution of H2 oxidation at the anode is
negligible. 82
Figure 4.5 Current under chopped illumination representing JOP for the PV–EC device in 0.5M KBi / 0.5M K2SO4 pH9.2. The chopped illumination illustrates the recovery in SFE and reproducibility in measuring JOP through the PV-‐EC device 83
Figure 4.6 Decay of the open–circuit voltage of the 4–cell PV mini–module over the course of ~15 min. The initial VOC at 2.42 V decays to a steady–state of 2.27 V after the first 10 min (▬▬, orange), which contributes to the initial decline in the SFE of the coupled PV–EC device. After overnight illumination, the Voc was measured (▬▬, blue) and shows a slight recovery to 2.31 V, which corresponds to the initial increase in SFE of the PV–EC device during the first 24 h. 84
Figure 4.7 Specific conductance measurements for various electrolytes considered to minimize RSOL. KOH (∎, red squares) is the most conductive electrolyte; in order to operate in pH near neutral regimes 0.5M KBi was used with additional supporting electrolyte, such as KNO3 (●, green circles) or K2SO4 (●, black circles). 85
Figure 4.8 Gas quantification for NiMoZn cathode operating in (left) 0.5 M KBi / K2SO4 and (right) 0.5 M KBi / KNO3 both at pH 9.2. The black line represents 100%
represent H2 measured by gas chromatography. The red arrow indicates when electrolysis was stopped. GC analysis was conducted until the moles of gas measured in the headspace reached a steady–state. The lag period (▬▬, black) in gas generation is due to the buildup of gases in the headspace of the EC cell. 86
Figure 4.9 Gas quantification for NiBi cathode operating in (left) 0.5 M KBi / K2SO4 and at pH 9.2. The black line represents 100% Faradaic efficiency based on the charge passes during electrolysis. The green circles represent O2 measured by gas chromatography. The red arrow indicates when electrolysis was stopped. GC analysis was conducted until the moles of gas measured in the headspace reached a steady–state. The lag period (▬▬, black) in gas generation is due to the buildup of
gases in the headspace of the EC cell. 87
Figure 4.10 Current under chopped illumination representing JOP for a PV–EC device composed of a 3–cell PV–module, a NiBi anode, and NiMoZn cathode operating in 1M KOH. Because KOH is a more conductive electrolyte, a 12% or greater SFE can be obtain with a 3–cell PV module as opposed to a 4–cell module. The initial drop in SFE is due to the decrease in PV efficiency, due to heating of the PV–module. The chopped illumination represents the recovery in SFE. 88
Figure 4.11 J–V curves of the individually measure PV and EC components making up the PV–EC device operating in 1M KOH. The grey curves represent the J–V curves for the PV modules composed of either three (
–––––
, grey–dashed) or four (▬▬, grey–solid) single–junction c–Si solar cells measure under AM 1.5 illumination. The blue curves represent electrochemical load J–V curves using NiBi and NiMoZn catalysts, where the ideal EC curve (–––––, blue–dashed) is based on previously reported Tafel analysis and the actual EC curve (▬▬, blue–solid) measured in a 2– electrode experiment. The point of intersection represents the JOP (●, orange circles)and the SFE of the coupled system. 89
Figure 4.12 SFE inferred from JOP for the PV–EC device operating in 0.5M KBi / 0.5M K2SO4 pH 9.2 measured for over 7 days of operation showing no decrease in SFE over operation time. Spikes are due to the addition of solution to maintain the
solution level and pH. 90
Figure 5.1 Tafel plot of a sputtered NiFeO OER catalyst operating in 0.5 M KBi / 1.5M KNO3 pH 9.2. A Tafel slope of 45 mV decade–1 is observed for a 50 nm (∎), 100 nm (●)and 200 nm (▲) thick NiFeO film. Inset: SEM image of a NiFeO shows a very
dense, compact film. 102
Figure 5.2 Tafel plots of 200 nm thick NiFeO (81% mol Ni, 19% mol Fe) on Ni-‐ coated glass operated in (▲) 0.2 M KPi, pH 7.0, 92 mV decade–1 slope; (■) 0.2 M KBi, pH 9.3, 61 mV decade–1 slope; (●) 1.0 M KOH, pH 13.9, 45 mV decade–1 slope. 103
Figure 5.3 Tafel analysis of Co–OEC films formed and operated in KBi (●) as
opposed to KPi (●) solution. The films formed from KBi exhibit a lower Tafel slope
and therefore demonstrate higher activity than those formed in KPi. 104
Figure 5.4 Tafel analysis of Co–OEC’s formed from anodizing metallic cobalt in KBi solution. In all cases the Co–OEC exhibits a Tafel slope of 60 mV decade–1, however starting with thicker metallic films produces Co–OEC’s with higher activity than
thinner films. 105
Figure 5.5 The current density traces show that recirculating streams allow the device to function stably and continuously (purple trace), while without recirculation the device performance deteriorates as concentration gradients form across the cell and ionic species are depleted in the oxygen-‐evolution side (red trace). The inset in the graph corresponds to a schematic representation of the parallel-‐plate solar-‐hydrogen generator. Reprinted with permission from reference
32. 107
List of Tables
Table 2.1. Summary of Faradaic efficiency for npp+–Si | interface| catalyst films 39 Table 3.1. Solar cell parameters for the modeling. 61 Table 3.2 Electrochemical parameters for the modeling. 61 Table 4.1. PV characteristics for the 3 and 4–cell c–Si mini–modules. 78
List of Abbreviations
ALD atomic–layer deposition AM air mass
a–Si amorphous silicon b Tafel slope
BJ buried junction BOS balance of systems Bi borate buffer cb bulk concentration
CIGS copper indium gallium diselenide
CoBi cobalt–based catalysts deposited from borate electrolyte CoPi cobalt–based catalysts deposited from phosphate electrolyte c–Si crystalline silicon
CV cyclic voltammogram or cyclic voltammetry CVD chemical vapor deposition
D diffusion coefficient
E-‐beam electron beam evaporation EC electrochemical
EF Fermi energy or level F Faraday’s constant FF fill factor
FTO fluorine doped tin oxide HEC hydrogen–evolution catalyst HER hydrogen–evolution reaction ITO tin–doped indium oxide J current–density
J0 exchange current density JSC short–circuit current kb Boltzmann’s constant MPP maximum power–point n ideality factor
NHE normal hydrogen electrode
NiBi nickel based catalyst deposited from borate electrolyte NiFeO nickel iron oxide
OEC oxygen evolution catalyst OER oxygen evolution reaction
P power
PCET proton–coupled electron tranfer PEC photoelectrochemical
PSII photosystem I PSII photosystem II Pi phosphate buffer PV photovoltaic q charge
R resistance s series
SEI semiconductor electrolyte interface SEM scanning electron micrograph SFE solar–to–fuels efficiency sh shunt
SJ solution junction sol solution
T temperature
TCO transparent conductive oxide th thermodynamic
VAppl potential applied to the electrode VOC open circuit voltage
WP watt at peak power η overpotential ηC coupling efficiency
ηEC electrochemical efficiency ηPV PV efficiency
δ Nernst diffusion layer
Acknowledgments
During the past six years at MIT and then at Harvard, I have had the pleasure of meeting some amazing people and scientists during this time and I would like to thank them for having an impact on me and my decisions.
From a scientific perspective I first have to thank my PhD advisor Dan Nocera. I have learned a lot from him. He has taught me many technical skills including how to write a paper, how to make my research accessible and interesting to others, and how to make pretty figures. His unique advising style of always pushing you when you need it has taught me more than anything else I’ve encountered in graduate school. He seems to always be so intuitive to what his students need to be successful and no matter how many times we mess up he never gives up on us.
I would also like to acknowledge my long–time collaborator Tonio Buonassisi for our many thought provoking meetings.
From my undergraduate academic experience I like to thank my under–graduate research advisor Dr. Stephen Mezyk for sparking my interest in doing research and pursuing a graduate degree.
I thank Dr. James Kiddle for being a great collaborator and kindred spirit. We have been great friends and have had so much fun together.
Now for the part most people skip to acknowledging all of the lab mates and colleagues that have inspired, influenced, and/or have just been great friends over the years:
I would like to thank Dr. Liz Young for telling me that I was not alone in feeling like I was the only person in my class who didn’t understand everything and felt way too behind to keep on going. I also admire Liz’s no nonsense attitude and the ability to always stand up for herself and others without being shy or afraid what others might think.
I will also have to thank Dr. Matt Kanan for always being so inspiring and kind even to a lowly first year graduate student. I was always so impressed seeing him at the Miracle of Science every Saturday with a new scientific paper to read along side a beer and burger. I also appreciate the friendship we have maintained over the years and how he always makes time to meet me for a drink when he is town.
Additionally I would like to thank Dr. Steve Reece. Although we didn’t overlap, being the great mentor that he is really helped me a lot during my first few years in
graduate school.
Dr. Mark Winkler was a great colleague and collaborator. For having so many helpful meetings and pep talks now and then.
Dr. Joep Pijpers got me started on my project and was a great mentor during the short time period we worked together.
Dr. Dino Villagran is one of the nicest people but somehow has made every female in lab cry over some ridiculous thing.
Dr. Alex Radosevich for teaching every one how to bootie bomb.
Dr. Bob McGuire for being such a fun and nice person.
Dr. Dilek Doğutan and I joined the Nocera lab around the same time. It has been nice seeing her progress from a post–doc to her current position where she has so much leadership responsibility. She really helps facilitate the research in our lab on a daily basis.
Dr. David Powers for giving great pep–talks during this last month and being a good friend.
Dr. Eric Bloch has been someone I have only known a short time but has been a really fun and kind person.
Dr. Tom Kempa for reading over portions of my thesis and for all of our long talks about science.
Dr. Chris Gagliardi for being such a nice and funny person.
Dr. Emily McClaurin for being such a good friend to me during my first few years. She always put up with my crisis (which were fairly often). She taught me a lot of things about how to handle myself in lab and our “CHEMREF” sessions were always helpful.
Dr. Changhoon Lee for being someone who I could never hear speak but I knew he was a kind person.
Dr. Yogesh Suredranath was the person I was most scared to present in front of at group meeting so I would go through my slides with him beforehand. He always gave time and attention to people who asked for it.
Dr. Matt Chambers for being the eternal optimist and always playing devil’s advocate. We had a lot of fun times especially at the Bleacher Bar. Let’s go Buffalo!
In my first year I started out as one of four and am the only one who made it through. I would specifically like to thank Pete Curtain for being such a smart and happy person. He is someone I have always missed, especially in times where I just wanted a person who could be a partner in crime during the various phases of graduate school. My last memory of a big hung before telling him good luck before he left is still one my favorite memories.
Kwabena Bediako for being so knowledgeable and helpful. However, sharing frustrations with science and graduate school with someone who makes it all seem so easy made me feel not so alone. I also always appreciate the pep talks walking home after a long day in lab.
Chris Lemon for being a great friend. Chris was always there when I needed him and was always ready to grab a beer and hang out after a long day in lab. He is one of the hardest workers in lab and never seems frustrated. I love our “gay–tes” at
Cambridge Common.
Andrew Ullman for being so quirky. I have loved seeing the transformation from hippie to clean–cut and dad–like (thanks Anne Marie). His love for reading old textbooks is hilarious and he has the best smile out of anyone in lab.
Mike Huynh for being the smartest, hardest–working, and kindest person in lab all of which comes completely naturally. I think I had the best person to give group meeting with and enjoyed his delicious home–cooked treats he would surprise me with.
Bon Jun Koo I don’t even know where to start. You have been a great friend and have always been there for me. Obviously my favorite thing about you is your confusion with the English language and American culture, which has made me laugh countless times. I also love your no nonsense attitude especially during long group meetings.
Nancy Li has been like a little sister to me over the past year. I love our talks about all things shopping and being terrible influences on one another when it comes to purchasing things we don’t need. She is so thoughtful and such a hard worker. I think she will have a very successful PhD experience.
Dan Graham all I can say is thank you for always being the scape–goat.
Bryce Anderson and Andrew Maher are both fun, sincere, and kind people and made sharing an office with no windows seem not so bad.
Evan Jones for always seeming to be in the wrong place at the wrong time, which makes me laugh.
Seung–Jun Hwang for putting up with all of our questions on the Korean language after Bon Jun confuses us.
There have been many people I didn’t get to know very well to all of you I wish the best of luck.
On a more personal note:
I would like to thank my amazing husband Eric Hontz. In the last two years he has helped me in every aspect of life. We have so much fun together and I am so excited about our future together.
I would like to thank my dad for always visiting me in every place I’ve lived and being proud of me.
I would like to thank my mother for being the strongest person I know. She has been so encouraging and helpful and I love her very much.
1.1 The need for clean–energy
One of the greatest challenges facing the world today is the need for clean– renewable energy resources to supply the needs of a quickly growing world– population. Current world energy consumption is 524 quadrillion BTU (5.5 x 1020
Joules or 17.5 TW per year).1 Due to an increase in world population to 3 billion
people by 2050, the world energy consumption is expected to increase by 56% and double by the end of the century. 1–3 Most of this population growth is occurring in
the developing world, which presently does not have the infrastructure or wealth to keep up with this demand.4
Presently 86% of the current world–energy is supplied by fossil fuels and it is projected that even with increase world population, fossil fuels can continue to power the planet for many years to come.5,6 However, increasing levels of CO2 in the
atmosphere have been rising since the industrial revolution when the world population was seven times less than today. Given that human activity led to increased concentrations of CO2 in the atmosphere with a considerably smaller
population, the impact of today’s rapidly growing world population could lead to much more severe results. The common goal amongst scientists and policy makers is to prevent the concentrations of CO2 in the atmosphere from reaching levels such
that the change in global temperature is more than 2oC.7 While it remains unclear
what impact the increased global temperature will have, it seems unwise to perform an uncontrolled experiment on the environment.
This quandary necessitates new technologies to produce and store
renewable energy that minimizes the environmental consequences associated with burning fossil fuels.
1.2 Renewable Energy
Due to the inefficiency of photosynthesis (1%)8 and the spatial limitations of
wind power,9 neither biomass nor wind is a viable option to fully meet the world
energy needs. The sun is by far the most abundant source of energy as more energy from the sun strikes the earth in just one hour than is presently consumed in one year. Impressively, covering 0.1% of the Earth’s surface with solar cells with an efficiency of 10% would satisfy present energy needs.10,11 Unfortunately due to the
intermittent and diurnal nature of sunlight, in order to make solar–energy as a viable resource requires capture, conversion, and storage.
Figure 1.1 Schematic showing (1) solar capture of solar energy by a photovoltaic device, (2) conversion of solar photons into a wireless current, and (3) storage via breaking the bonds of H2O to make H2 which can be used as a fuel. Adapted from ref.
1.3 Capture of solar power and conversion to electrical power
An elegant technological approach to directly convert sunlight into electricity without moving parts or environmental emissions is to utilize semiconductors. Semiconductors take advantage of the fact that photons with energy equal to the optical band–gap (similar to HOMO–LUMO transition for molecules) can create an electron–hole pair that can be separated between two different materials, thus effectively establishing a potential difference across the interface. However, since semiconductors are transparent to photons below the band–gap and photons having energies much higher than the band gap rapidly release heat to the lattice of the solid the upper bound conversion efficiency of solar power input to electric power output of a single–absorber is 32% based on a semiconductor with a band–gap of 1.4 eV. 12
Figure 1.2 Solar irradiance at the surface of the Earth. The band–gap of silicon is overlaid as an example showing that photons absorbed at the band–gap can be converted and those absorbed above the band–gap are wasted as heat.
After photogenerated electrons and holes are created, an electric field is required to separate charges such that they can be transferred to an external load. An electric field can be established by interfacing a semiconductor with another material containing a different work function (also called Fermi level, electron affinity). This can include a metal, another semiconductor, doping two sides of the same semiconductor, or an electrolyte containing a redox couple. Once interfaced, charge transfer between the two materials occurs until equilibrium is established. This produces a region in each material that is depleted of majority charge carriers (electrons for an n–type semiconductor and holes for a p–type semiconductor), which is depicted as band–bending within the semiconductor (upward for n–type, downward for p–type). This translates to a built in potential due the electric field formed at the junction. Upon illumination, a non–equilibrium concentration of photogenerated electrons and holes disturb the previously established equilibrium formed at the interface and the electric field serves to separate the photogenerated electrons and holes such that they can be extracted to do electrical work. The electrical power generated could be used directly. However due to the intermittent nature of sunlight, it is also important to store the electrical power generated in a fuel.
1.4 Conversion of electrical power into fuels
The best–known example of converting solar energy and storing it as
chemical energy can be found in nature. Photosynthetic organisms capture sunlight and convert water and carbon dioxide into oxygen and reduced organic species,
due to the high energy density the chemical bond. The primary steps in
photosynthesis are absorption of solar energy by chlorophyll and other pigments, after which the photogenerated electrons and holes are separated in the
Photosystem II (PSII) reaction center. The oxidative power of the photogenerated holes in PSII are transferred to the oxygen evolving complex to split water,
producing molecular oxygen which is released into the atmosphere, as well as protons and electrons which are transferred and consumed in Photosystem I (PSI) to reduce NADP+ into NADPH (natures form of hydrogen), which is ultimately used
to reduce CO2 to carbohydrates. Since products from the water–splitting reaction
are subsumed in subsequent photosynthetic processes, water–splitting is the most critical step in photosynthesis.6,13,14
The thermodynamics of water–splitting can be described by the following oxygen evolution and hydrogen evolution electrochemical half reactions (OER and HER, respectively): 2𝐻!𝑂 → 𝑂!+4𝐻!+4𝑒! Eoanode = 1.23V − 0.059(pH) vs. NHE (1.1) 4𝐻!+4𝑒! →2𝐻 ! Eocathode = 0V − 0.059(pH) vs. NHE (1.2)
combining equations (1) and (2) indicates that a total voltage of 1.23 V is required to drive the uphill water–splitting reaction. However, additional voltage is necessary to drive the reaction kinetics or rate of the reaction for a given current density (JEC)
making the overall voltage for water–splitting (VEC):
𝑉!" 𝐽!" = 𝜂!!+ 𝜂!"# 𝐽!" + 𝜂!"# 𝐽!" + 𝜂!(𝐽!") (1.3)
where, ηOER and ηHER are the anodic and cathodic overpotentials, respectively, that
arise from the intrinsic activation barrier for the electrochemical half–reaction occurring at the electrode–solution interface and ηR accounts for resistive losses
which can arise from resistance through the electrodes, contacts, or mass transport limitations. Water–splitting catalysts can minimize ηOER and ηHER. While the impact
of ηR can be minimized through optimal cell designs,15–17 the activation
overpotentials are intrinsic properties of the catalysts utilized at the anode and cathode. This overpotential, which is also a metric for catalyst activity, is typically reported in units of mV decade–1, and is logarithmically related to the current
density (J) as given by the Tafel law18:
𝐽= 𝑏 log 𝐽𝐽
! (1.4)
where b is the Tafel slope and J0 is the exchange current–density that characterizes
the intrinsic activity of the electrode under equilibrium conditions. In order to optimize the efficiency for water–splitting, that is the ratio of the thermodynamic potential for water splitting to the thermodynamic potential, catalysts exhibiting high J0 and a low Tafel slopes are necessary.
1.5 Photoelectrochemical water–splitting
The concept of a photoelectrochemical (PEC) device was first popularized by the 1976 paper of Fujishima and Honda. 19 They described immersing a TiO2
semiconductor in solution, illuminating it with UV light, and observing upon
application of a potential bias the evolution of both hydrogen and oxygen. Since this study, hundreds of device–constructs have been investigated as PECS. They can broadly be classified as those that either employ a solution junction (SJ) or buried junction (BJ) for charge separation.20,21,22 While the physical principles underlying
the operation of the methods are quite similar,21 the position of charge separating
interfaces relative to interfaces injecting charge into water redox couples has important consequences for implementation of either method. SJ–PEC operates on the principle that upon submerging a semiconductor in a solution containing a redox couple, charge transfer at the interface will occur provided appropriate alignment between the semiconductor Fermi level (EF) and the Nernst potential of
redox species. The depletion region formed due to band–bending within the semiconductor allows for charge injection into the solution. For the case of water splitting by a SJ–PEC, the quasi–Fermi level for photogenerated electrons or holes must straddle the thermodynamic potential for the water–splitting reaction (i.e. 1.23 V).23 Due to the previously mentioned kinetic overpotentials the actual voltage
required for water–splitting lies between 1.6–2 V. Since the photovoltage generated from a semiconductors is typically at least 0.4 V less than its band–gap,24 this
requires the semiconductor to have a band–gap in excess of 2 V. Therefore, even with proper band–alignment, only a small fraction of the solar spectrum can be
utilized limiting the efficiency to 7%.23,25 Furthermore, semiconductors are rarely
good water–splitting catalysts.26 This limitation may be addressed by depositing
water–splitting catalysts on the semi–conductor surface. But surface modification often affects the efficiency of light absorption and charge separation through the semiconductor–electrolyte interface (SEI).2728 Since charge separation and catalysis
are intimately tied together, optimization of optimization of the individual components of such a device is challenging and such devices have only demonstrated solar–to–fuel efficiencies (SFE) of less than 1%.29
1.5.1 BJ–PEC requirements
Many of the aforementioned challenges with SJ–PEC can be overcome by relying on a solid–state semiconductor–semiconductor junction (also referred to as a buried junction) to perform charge separation. In the BJ–PEC configuration a
Figure 1.3 Qualitative schematic of an n–type semiconductor/electrolyte junction for photoelectrochemical water–splitting.
solid–state junction is formed either between two semiconductors or by doping two sides of the same semiconductor. By controlling the doping–levels, the width of the depletion region can be optimized for maximum charge separation.30 Thirty years of
ongoing research in the photovoltaic (PV) community has led to doping as a mature technology and optimal charge separation and photovoltage characteristics has been achieved.31 The buried–junction can be connected to relevant interfaces (e.g.
for charge injection to catalysts) through Ohmic contacts, which can be either thin– metal films or conductive oxides deposited on the surface of the semiconductor. Since the semiconductor surface is completely protected from the aqueous environment, semiconductor stability no longer poses a problem. Ultimately, the only requirement is of the BJ–PEC is that an appropriate voltage is supplied to drive the HER and OER conversions.29
Figure 1.4 Qualitative schematic of a buried semiconductor/electrolyte junction for photoelectrochemical water–splitting.
Furthermore, decoupling the absorption and charge rectification properties from the water–splitting catalysis enables independent optimization of all the required components. The Ohmic contacts can be optimized by choosing highly conductive materials with proper band–alignment to allow facile charge transport.32,33 Water–
splitting catalysts can be independently evaluated and interfaced.
1.5.2 BJ–PEC devices
Many buried junction BJ–PEC devices have been demonstrated in the last 30 years. To date the highest solar–to–fuels efficiency (SFE) devices utilized either expensive multi–junction III–V solar cells,34,35 low efficiency amorphous silicon (a–
Si) solar cells,35–38 and most recently copper indium gallium diselenide (CIGS) solar
cells.39,40 In all cases the integrated BJ–PEC device suffered from either low SFE,36–38
and/or were composed of expensive PV materials, expensive catalysts, and operated in strongly acidic or basic electrolytes hindering long–term stability.34,35,39,40 For
these reasons, none of these devices were realistic for economic viability. In order to make this technology realistic from both a cost and stability perspective, low–cost high efficiency PV materials and high efficiency earth–abundant catalyst that operate in benign aqueous environments are necessary.
1.6 Crystalline Silicon
Silicon is prime candidate material for buried–junction devices owing to its almost optimal band–gap of 1.1 eV which absorbs a large fraction of the solar spectrum and it is the second most abundant material on the planet. Additionally
silicon solar cells and modules are one of the most mature technologies developed for solar capture and conversion.31 Currently, the record solar conversion efficiency
for c–Si solar cells has hit 25%, which is quite impressive considering the
thermodynamic limit of 29%.31,41,42 Traditionally silicon PV’s have been thought to
be too expensive.11,13,43,44 However, after 30 years of optimization the price of silicon
solar cells has declined and the conversion efficiency has improved. 31,41 ,45 From
2004–2008 crystalline silicon (c–Si) PV modules remained steady at $3.5–$4 per peak watt (WP–1). However, due to the price decrease in polycrystalline silicon,
which is used a feedstock material for c–Si, in 2008 the price decreased by half and in 2011 fell below $1 Wp–1.41–47 In order to be cost competitive with current the
baseload fossil fuel electrical utility plants in the US without subsidies the price
needs to further decrease to $0.5–0.75 WP–1. Modeling and outlined pathways show
Figure 1.5 Chinese c–Si PV module prices since 2006. The data was adapted from ref. 43.
that this goal should be achievable by the year 2020.47,48 However, even the current
status of c–Si PV’s has made them a cost–competitive technology with the current resources used in developing nations such as Africa, the Persian Gulf, and India.45
1.7 Earth–abundant water–splitting catalysts
Traditionally catalysts for water–splitting include rare earth elements of noble metals including Pt, Ir, Ru.49–51 Our labs changed the paradigm by discovering
active catalysts composed of Earth–abundant materials. Oxidation of Co2+ salts in
buffered solutions yield a cobalt–oxide water–oxidation catalyst self–assembles onto conductive substrates.52,53 This technique has been extended to other earth–
abundant metals such as Ni and Mn.54–56 These catalyst are stable by virtue of a self–
healing mechanism,57–60 and they operate under a variety of pH ranges,55,56,61,62 and
in the presence of impurities. 61,62 Additionally, it has been shown that these
catalysts can be easily interfaced with semiconductors63–68 and specifically with
buried–junction silicon PV’s.36,69–71 Since these OER catalysts operate under a variety
of pH neutral conditions, the choice of catalyst for the hydrogen evolution reaction (HER) has not required platinum.36,72 Specifically, NiMo(Zn) alloys for hydrogen
evolution, which also self–assemble onto conductive substrates from an aqueous solution containing Ni2+, sodium molybdate and anhydrous zinc chloride in the
presence of pyrophosphate, bicarbonate, and hydrazine. Subsequent leaching in base produces a high surface area material.73 Theses alloys are able to achieve
leaching, can attain activities as high as at 1000 mA cm–2 at an overpotential of 35
mV. 72,74
1.8 Overview
The following chapters of the thesis will discuss the interfacing of water– splitting catalysts with c–Si photovoltaics to produce BJ–PEC devices using a completely modular approach. Chapter 2 focuses on directly depositing OER catalysts onto single–junction c–Si PV’s to create a light–assisted photoanode. Of particular importance is the ability to protect silicon from the oxidizing conditions required for water–splitting with a protective interface. Fabrication of these silicon photoanodes requires optimization of two interfaces: a silicon–protective layer interface and a protective layer catalyst interface. Optimization of both lead to a lower overpotential (as determined by Tafel analysis) required for OER.
Since a single–junction c–Si solar cell does not supply the voltage required to achieve water–splitting without the use of an external potential bias, in order to realize a stand–alone water–splitting device based on c–Si, multiple single–junction
Figure 1.6 Depictions of the molecular structure of our Mn, Co, and Ni water– oxidation catalysts. Reprinted with permission from Mike Huynh.
c–Si solar cells need to be connected in series. Given that the technical aspects of device integration can be quite challenging, it is beneficial to predict the behavior of a coupled photovoltaic–electrochemical device (PV–EC). In Chapter 3, steady–state equivalent–circuit analysis of a PV based on a string of single–junction c–Si solar cells driving an electrochemical load based on the OER–catalysts developed in our lab allows us to predict the coupled behavior between the PV and EC components. Importantly this allows us to observe the impact solar–to–fuel efficiency based on parameters such as choice of catalysts as well as resistive losses.
Guided by modeling and simulation, a modular PV–EC device is presented in Chapter 4 that is constructed from c–Si and non–precious catalysts. A 10% solar–to– fuels efficiency is demonstrated. This chapter illustrates how a modular approach allows for independent characterization of each component.
The fi
